Comparison of Estimates of Solar Radiation from Neural Network Models and Angstrom Formula for Application in ET0 - A Case Study in Abureyhan Station, Iran

Abstract

Solar radiation (RS) data are desirable for calculating reference evapotranspiration by Penman-Monteith (PM) method. In the absence of actual RS data, the Angstrom equation has been recommended by the Food and Agriculture Organization of the United Nations (FAO). This equation requires actual Sunshine duration that is not commonly observed in many weather stations. This article presents the outcome of an attempt made to predict the RS based on measured values of maximum and minimum air temperature only. This is important because temperature data are commonly available parameters. Data for Abourayhan station, southwest of Tehran, were used for training and evaluating a feedforward ANN using backpropagation algorithm and calibration of Angstrom equation. The study demonstrated that modelling of daily RS through the use of Angstrom equation gave better estimates than the ANN technique. RMSE and R2 for the comparison between observed and estimated RS for the tested data using the proposed ANN model are 2.41 MJ m-2 d-1 and 0.88, respectively. For the Angstrom method these values are 1.42 MJ m-2 d-1 and 0.96. Our analyses also indicate that differences in daily RS between the different procedures have less effect on estimated ET0 by using PM equation.